"Strength does not come from physical capacity. It comes from an indomitable will" Gandhi On to the next one. Keep focus.

Demand is accelerating 🍻 Marvell read through looks really compelling. $SMTC $MRVL

High conviction for 5/26 earnings: $SMTC $12.5B market cap. Here is the thesis: 1. The 1.6T and Trainium Volume Ramp: Semtech completely dominates the TIA space, meaning massive content per box growth as 1.6T scales literally right now. Throw in the colossal scale out Ethernet topologies AWS needs for Trainium. The sheer volume of optical connections required feeds directly into their bottom line. 2. The Vera Rubin 224G Catalyst: Inside the rack copper is still king. When you push 224G over standard copper wire, the signal completely dies after about two meters. It is a hard wall. If hyperscalers want to deploy Rubin without literally melting their power grids or dropping data packets, passive copper is entirely dead to them. They are structurally forced to buy Active Copper Cables or shift entirely to Linear Pluggable Optics. 3. The Agentic AI Inference Boom: Agentic AI is forcing a massive build out of inference clusters that are ridiculously sensitive to latency and power draw. Whether these racks run on GPUs or custom CPUs, they demand networking that is fast and highly efficient. By stripping out power hungry digital chips, Semtech becomes the ultimate hardware agnostic arms dealer for next gen networks. Semtech wins the edge. I also like the smart money here with Atreides increasing its stake. NFA.

$LITE beat today confirms the read-through on Silicon Photonics demand. ⚡️ $TSEM flying under the radar as the only scalable Western foundry for the AI interconnect ramp (via Intel NM). "Situational Awareness" supply chain hedge. 1-Year Chart: +170% (quietest multi-bagger in semi). Earnings: Feb 11.

I don't have too too much to add on top of this earlier post on V3 and I think it applies to R1 too (which is the more recent, thinking equivalent). I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed in AI. You may not always be utilizing it fully but I would never bet against compute as the upper bound for achievable intelligence in the long run. Not just for an individual final training run, but also for the entire innovation / experimentation engine that silently underlies all the algorithmic innovations. Data has historically been seen as a separate category from compute, but even data is downstream of compute to a large extent - you can spend compute to create data. Tons of it. You've heard this called synthetic data generation, but less obviously, there is a very deep connection (equivalence even) between "synthetic data generation" and "reinforcement learning". In the trial-and-error learning process in RL, the "trial" is model generating (synthetic) data, which it then learns from based on the "error" (/reward). Conversely, when you generate synthetic data and then rank or filter it in any way, your filter is straight up equivalent to a 0-1 advantage function - congrats you're doing crappy RL. Last thought. Not sure if this is obvious. There are two major types of learning, in both children and in deep learning. There is 1) imitation learning (watch and repeat, i.e. pretraining, supervised finetuning), and 2) trial-and-error learning (reinforcement learning). My favorite simple example is AlphaGo - 1) is learning by imitating expert players, 2) is reinforcement learning to win the game. Almost every single shocking result of deep learning, and the source of all *magic* is always 2. 2 is significantly significantly more powerful. 2 is what surprises you. 2 is when the paddle learns to hit the ball behind the blocks in Breakout. 2 is when AlphaGo beats even Lee Sedol. And 2 is the "aha moment" when the DeepSeek (or o1 etc.) discovers that it works well to re-evaluate your assumptions, backtrack, try something else, etc. It's the solving strategies you see this model use in its chain of thought. It's how it goes back and forth thinking to itself. These thoughts are *emergent* (!!!) and this is actually seriously incredible, impressive and new (as in publicly available and documented etc.). The model could never learn this with 1 (by imitation), because the cognition of the model and the cognition of the human labeler is different. The human would never know to correctly annotate these kinds of solving strategies and what they should even look like. They have to be discovered during reinforcement learning as empirically and statistically useful towards a final outcome. (Last last thought/reference this time for real is that RL is powerful but RLHF is not. RLHF is not RL. I have a separate rant on that in an earlier tweet xitter.cf/karpathy/status/182127…)